Quasar Continuum Fitting and Silicon Absorption
in the Low Redshift Intergalatic Medium
by
Adam A. Miller
Submitted to the Department of Physics
in partial fulfillment of the requirements for the degree of
Bachelor of Science
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
May 2006-
® Adam A. Miller, MMVI. All rights reserved.
The author hereby grants to MIT permission to reproduce and
distribute publicly paper and electronic copies of this thesis document
in whole or in part.
7ii7E
9
3RARI%
Author ................................................
........
Department of Physics
May 12, 2006
Certified
by...................................-W......
..
..
Scott Burles
Assistant Professor
Thesis Supervisor
Acceptedby.......................
,.........-..
......
Professor David E. Pritchard
Senior Thesis Coordinator, Department of Physics
ARCHIVES
Quasar Continuum Fitting and Silicon Absorption in the
Low Redshift Intergalatic Medium
by
Adam A. Miller
Submitted to the Department of Physics
on May 12, 2006, in partial fulfillment of the
requirements for the degree of
Bachelor of Science
Abstract
We present results on the evolution of Lyae absorption at low redshift, and the first
systematic search for Si II absorption systems in the low redshift IGM. Our sample
consists of 832 Lyca absorbers from 328 spectra of 204 QSOs taken from the Hubble
Space Telescope archive. We develop a new, reproducible method of quasar continuum
fitting, designed to quickly identify absorption lines and measure the relative line
strength (a proxy for equivalent width). Our method, which fails to identify the
weakest lines, does manage to detect the strong features in a given spectrum and
provides enough information to identify metal absorption line systems. We confirm
the results of previous studies of Lya evolution at low redshift and find the number
density of absorbers can be described by a power law in (l+z) that is much flatter
than that found for Lya evolution at high redshift. Specifically, we measure a power
law index of -y = 0.57 ± 0.16 for lines with a rest relative line strength greater than
0.10 A. We also identify the presence of 14 Si IIsystems at z - 1. The number of Si
ii 1193 and 1260 A systems per unit redshift path length at a mean redshift of z =
0.9 is < N(z) >= 1.6 ± 0.6. This density is similar to that found for C Ii, Mg ii, and
O vi absorbers.
Thesis Supervisor: Scott Burles
Title: Assistant Professor
3
4
Acknowledgments
I would like to thank Scott Buries for his comments, suggestions, and discussions
about the content of this paper. And for putting up with me...
5
6
Contents
1 Introduction
13
2 Data
17
2.1
The Sample ................................
17
2.2
Reduction .................................
18
3 Continuum Fitting Technique
3.1
4
Line Identification and Equivalent Width Measurements
.......
Results
22
25
4.1
Comparing the Smoothing Method with Hand Fit Continua
4.2
Evolution of Lya Absorbers at z < 1.7 .................
27
4.2.1
28
4.3
5
19
.....
The Cumulative Distribution as a Function of Redshift ....
Si II Detection in the IGM ........................
4.3.1
Search for Si II Lines.
........
4.3.2
The Distribution of Si II .....................
Conclusion
25
31
..............
32
33
41
7
8
List of Figures
3-1
Fitted Continuum Spectra ........................
20
3-2
Fitted Continuum Spectra ........................
21
3-3
RS Comparison of Two Smoothing Kernels ...............
23
4-1
RS vs. Equivalent Width
26
4-2
Lya Forest Lines .............................
4-3
Cumulative Distribution of Lyac Lines vs. Redshift
4-4
Cumulative Distribution of Si ii Lines vs. Redshift ..........
........................
9
29
..........
30
.
38
10
List of Tables
4.1
Si ii System Line List.
.........................
11
33
12
Chapter
1
Introduction
A particularly useful feature of quasi-stellar objects (QSOs) is that they provide a
luminous background source to probe the intergalactic medium (IGM). Of particular
interest is the Lya forest where the majority of observable intergalactic H gas resides.
Investigating the distribution and evolution of hydrogen throughout the universe is
vital towards understanding and creating models of galactic structure and formation.
Several previous studies have characterized Lyac absorption at high redshift [1], [2],
however, optical spectra are limited in the amount of information they can provide
about low redshift Lya absorption. From the ground, Lyce emission and absorption
can only be observed at z > -2. At lower redshifts Lyca becomes obscured by atmospheric interference. Therefore, to study Lyaoabsorption at z < 2 requires the ability
to probe and the use of ultraviolet spectra.
Fortunately, the Hubble Space Telescope (HST) Faint Object Spectrograph (FOS),
Goddard High Resolution Spectrograph (GHRS), and Space Telescope Imaging Spectrograph (STIS) are ideal for such studies of low redshift Lya systems. Previously,
a large survey identifying all H and metal absorption lines has been conducted for
all FOS QSO spectra [3], while the evolution of Lya absorption lines at low redshift
is also well studied
[4], [5], [6], [7]. These projects are limited, however, in that they
only include spectra from FOS. Each of these studies employs the use of hand fit continua to identify absorption lines in the IGM. While a well established method, hand
fit continua are slightly problematic in that the method is not entirely reproducible.
13
Therefore, the possibility exists that two different people analyzing the same data in
the same manner could still produce differing results. Hand fitting continua can also
be a very laborious and time intensive practice. The development of a reproducible,
computational method of continuum fitting would serve to unify the way in which
Lya absorption is measured, while at the same time drastically reducing the amount
of time necessary to determine the continuum.
One particularly interesting difference between the ground-based and space-based
observations of the Lya forest is the dissimilarity in it's properties in the two different
regimes. The distribution of Lya absorbers in redshift, z, can be characterized by a
power law in (l+z). However, there seems to be a sharp break in the power law index
at z-
1.8, with Lu, Wolfe, & Turnshek [1], and Bechtold [2] measuring a power law
index of -1.7 for z > 2, while Weymann et al. [6], and Dobrzycki et al. [7] measure
a power law index around 0.5 for z < 1.7. Characterizing the exact distribution of
hydrogen over the history of the universe is vital to our understanding of cosmological
evolution.
The presence of metal. (anything that isn't H or He) absorption systems in the
Lya forest is well documented, however, there have been few direct searches for
such systems. Formed in the cores of stars and later exploded into the surrounding
IGM following the death of a star, metals can serve as possible tracers of supernovae
by answering questions about when and where they occurred. A major problem in
identifying metal gas clouds is that several of the stronger absorption lines are located
in the Lya forest where line blending and misidentification are significant difficulties
in determining metal absorption line systems. Some of these issues are addressed by
Buries & Tytler [8], in their search for O vI. Previous searches for Mg ii [9], and C
IV [4] have also been made. However, there have been no systematic searches for Si
in the IGM, one of the most common elements in the universe after hydrogen and
helium, despite several studies that have identified the presence of silicon [4], [5], [8],
[6], and
[3].
In this paper we expand upon the work of [7] to include all QSO observations
taken with FOS, GHRS, and STIS. The goal is to generate the largest sample of low
14
redshift Lyr absorption lines to date. We develop a new, reproducible method for
fitting the continuum of QSO spectra. In § 2 we present the data sample, along with
the criteria for selection, and the reduction process. In § 3 we introduce our method
of continuum fitting and line identification. We compare the method with hand-fit
continua in § 4. Using the line list from § 3 we also measure and discuss the evolution
of both Lya. and Si ii at low redshift in § 4.
15
16
Chapter 2
Data
2.1
The Sample
Our sample consists of all known QSO spectra with z > 0.33 taken by HST FOS [10],
GHRS [11], [12], and STIS [13]. We extend the sample of [14] to include all public
data available from the HST STIS archive. The redshift limit was chosen such that
the intrinsic Lyman limit (912 A) is viewable by HST. Some QSOs were removed or
omitted from our sample based on the following criteria:
1. Object is a known or suspected broad absorption line QSO.
2. Spectrum is high-resolution with small wavelength coverage (medium and highres modes on both GHRS and STIS).
3. Low signal to noise (approximately < 1 pixel- 1 ).
4.
Zem
>
1.8.1
Also, if two spectra in the sample provide almost identical wavelength coverage for example, if both FOS G130H and GHRS G140L data are available - then only the
higher quality spectrum is used. We include both spectra, however, in cases where
there is only partial overlap - for instance, FOS G270H and FOS G160L.
'We elect to remove QSOs with a redshift greater than 1.8 because in this regime ground based
optical observations provide better resolution and higher S/N of Lya than HST.
17
After the selection criteria have been applied, our sample contains 328 different
spectra of 204 QSOs.
2.2
Reduction
Before the continua were fit, each spectrum was processed by a series of steps described
below:
1. The spectra were corrected for Galactic extinction.
2. Strong features not intrinsic to the QSO or absorption clouds are masked out,
including instrumental artifacts, airglow, and cosmic rays.
3. The data are shifted into the QSO rest frame for continuum fitting.
Provided the correction for Galactic extinction comes before the data are shifted into
the object's rest frame the order of the previous steps is unimportant.
Below we
discuss the steps in slightly more detail.
The Galactic extinction correction is based on the analytical form of [15] using
E(B - V) values from [16]. We assume that Rv = 3.1 for all QSOs.
The features that were not intrinsic to the QSO were masked by removing the
individual pixels such that the code which performs the continuum fit ignores the
bad pixels. If the masked pixels lied below the continuum following the fit then they
were also masked from the search for absorption lines to prevent the false positive
identification of non-lines.
18
Chapter 3
Continuum Fitting Technique
We introduce a new method of continuum fitting whereby the continuum at each
pixel is determined by the weighted mean of the flux at that pixel and its nearest
neighbors. The continuum is the convolution of the flux with a smoothing kernel,
which is designed to be a half sine wave:
m-1
Ci= E
j=o
where Ci is the continuum at the
ith
Fi+jrm/2
Kj
(3.1)
pixel, F is the flux, K is the smoothing kernel,
and m is the total number of elements in the kernel. The kernel is normalized to
unity and arranged such that the maximum amplitude of the half sine wave is always
centered on the it h pixel. At the edge of each spectrum the kernel is truncated such
that only pixels which overlap the kernel are included in the sum. We employ the
use of two smoothing kernels, which are identical in every feature except their width.
The kernels were chosen to have a base width of 5000 km s-' and 10,000 km s-
at
Lya, or roughly 20 and 40 A, respectively.
In Figures 3-1 and 3-2, we show 6 spectra and the fits produced by our method
with the 5000 km s - 1 smoothing kernel. We choose to only include the smaller of
the two smoothing kernels to improve the clarity of the figures. The sample of QSOs
shown is chosen to demonstrate the characteristics of the fits over a wide range in
both QSO redshift and observed wavelength. We have also chosen 6 different HST
19
1253-0531
-
FOS G130H
_
14
Ze.
IIIIIIII11IIIIII I
IIII
Ill..'I I.
., III
I,,
I 1..1 I.1. I I I I I
= 0.536
I IL .1.1
6
4
EIl
II ". II. J
1250
1200
- I- I I
~~~~~~~
4~~~~~~~~~~'
1300
- "
I -
, I
-1 I
... 11
II-
I
J.-I
I I
. . . . . .
1350
1400
1450
1250
100
1550
1600
o<~
E
o
(1
n
1650
0
1700
1750
1800
1850
1900
1950
2000
2050
2100
2150
2200
2250
2300
A<
LL:
2300
2400
2500
2600
2700
2800
2900
3000
3100
3200
Wavelength (A)
Figure 3-1: Spectra from 3 different QSOs plotted with the 5000 km s-1 kernel fitted
continuum (dashed line) and 1 a error.
gratings with a variety of intrinsic QSO features, especially Lya and Ly/3 emission.
We note that there are some limitations to this new method. While the smooth
continuum will serve as a very good approximation of the "true" continuum in regions
where there is no absorption or emission, the accuracy will suffer near strong features
in the spectrum.
By definition the continuum will always be underestimated near
any absorption features (especially strong lines), because the reduced flux in these
areas will pull the weighted mean down. For instance, see the strong features at
1215 A in 2112+0555 and
2660
A in
0126-0105 (Figure 3-2, and notice that the
continuum is severely underestimated near these features. Conversely, in the vicinity
of any strong emission features, which are intrinsic to all QSO spectra, our method
will underestimate the true continuum at the peak of the feature and overestimate
the true continuum on the edges, or wings, of the emission profile. This can be seen
20
l 3UU
.
- - -.
140U
. --
---
10U
1U00U
---
1IUU
---
1UU
- -IUU
- -2Uo0
-- -
2100
-- -
2200
--
--2300
2400
--2500
·Q
(N
1
V) 1C
n
cn
0
1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
ILL
1700
1900
2100
2300
Wavelength
2500
2700
2900
3100
(A)
Figure 3-2: Spectra from 3 different QSOs plotted with the 5000 km s- l kernel fitted
continuum (dashed line) and 1 a error.
near the Ly&oand N v emission features (- 2960 A) in the spectrum of 1026-0045
(Figure 3-1).
It is possible to reduce these effects by changing the size of the kernel, however.
A very wide smoothing kernel is less susceptible the effects of smaller absorption
features, but at the same time it will do a terrible job of predicting the continuum
near emission features. A narrow kernel would do a much better job fitting emission
features, but it would also be extremely sensitive near absorption lines. It could
potentially underestimate the continuum enough that real absorption lines no longer
pass a significance test. In an effort to minimize the above effects we have elected
to use two kernels of moderate width, as mentioned above. Despite the limitations
mentioned above, our method is still employable for the identification of absorption
lines, especially strong metal systems.
21
3.1
Line Identification and Equivalent Width Mea-
surements
Using the method described above it would be impossible to accurately measure the
equivalent width of any absorption lines in the spectra because our method does not
replicate the "true" QSO continuum.
Therefore we define a proxy for equivalent
width - the relative line strength (RS), which is the relative absorption of a three
pixel window at each pixel.
j=i+l (Cj - Fj)
RS = E
j=i-1
C F
C
,
(3.2)
where RSi is the relative line strength of the it h pixel and 6Aj is the width in A of
each pixel. We note that in the case of saturation the equivalent width will approach
an upper limit equal to the size of the three pixel window. The uncertainty in RSi is
defined as the error in Fi for each pixel normalized to the value of the the continuum
and added in quadrature. We define the significance level, SL, to be RS/a(RS),
and
consider any pixel with SL > 3 to be a candidate for a line. We lastly require all pixels
with SL > 3 to be more significant than their two neighboring pixels on either side to
qualify as an absorption line. The line centroids are measured by taking the weighted
mean of the wavelength (A) at the three pixels, where the weights are equal to the
SL2 at that pixel. While line centroids are typically determined using a gaussian fit to
the absorption profile, this approach is not appropriate here because the continuum
is underestimated. Thus, the assumption of gaussian absorption profiles is a bad one
given that the continuum is underestimated at the wings of each absorption profile.
After applying these steps on every spectrum in our sample, we find that the two
smoothing kernels measure approximately the same RS. Figure 3-3 shows the RS for
the 10000 km s-l filter vs. the RS for the 5000 km s- 1 filter. Due to the difficulties
of measuring the continuum near emission lines, as described above, we reject any
lines near the two strongest emission features in QSO spectra: Lya and C v 1549 A.
Based on the average strength of each feature we reject any lines that fall in the range
22
..
....
11
4
-
Q9
1
o
2
oo
o
0
·.
::
::
1
.·r
::
·:
:iu·
:
I
0
- I
I
0
I
I
I
I
I
I
I
I
I
i
I
I
I
L I
I
I
I
I
I
1
2
I
I
I
I
I
L I
I
I
I
I
(
3
I
(
(
I
I
I
I
i
I
I
I
I -
4
5000 km s - 1 Kernal RS (A)
Figure 3-3: Scatter plot of RS as it is measured by both smoothing kernels. Points
where RS = 0 indicate a line detected by only one kernel, as described in the text.
The line represents the best fit to the points where RS > 0 for both kernels. It has
a slope of 1.059.
1190 - 1250 A or 1530 - 1570 A in the rest frame of the QSO. We chose an asymmetric
range about Lyo because N v 1240 A,typically a strong feature when present, blends
with the red side of Lya and prevents a reliable measure of the continuum by our
method until roughly AQSO = 1250 A.
Following the above removal, we measure a total of 2157 absorption lines with both
filters. The best fit line to these data yields a slope of 1.059 (RS 1 0000 = 1.059 x RS5000),
meaning the two methods measure roughly the same RS. The slightly higher RS for
lines measured by the 10000 km sl
kernel is to be expected given that this filter
23
averages over a larger range in A and therefore will not be underestimated as much as
the other filter at the location of the absorber. In addition to the 2157 lines measured
by both filters there are 422 lines that were measured only by the 10000 km s- 1 filter
and 134 lines measured only by the 5000 km s- 1 filter. These lines are shown in
Figure 3-3 as having a reduced line strength of zero for the filter in which they were
not detected. As can be seen from Figure 3-3 the majority of these are weak lines
with 83.5% having a RS < 1 A. Because the 5000 km s- 1 filter is more selective, and
therefore more likely to identify a higher percentage of real absorption features, we
elect to use its line list for all further analysis in this paper.
24
Chapter 4
Results
4.1
Comparing the Smoothing Method with Hand
Fit Continua
Measuring the continuum of a quasar with the method described in §3 is only useful
if one has a sense of how the RS and the line lists are related to their analogous
measurements made with the "true" continuum. Fortunately, Bechtold et al. [3] fit
by hand the continua to all QSO spectra taken with FOS G130H, FOS G190H, and
FOS G270H. Our sample and the sample of Bechtold et al. [3] have 159 spectra of
108 QSOs in common, this sample is large enough to compare our method with the
hand fit method.
Figure 4.1 shows the observed RS vs. equivalent width as it was measured by
Bechtold et al. [3]. The line represents the best fit to the data and has a slope of
0.378. Thus, assuming Bechtold et al. [3] fit the true continuum, an approximate
conversion from RS to equivalent width can be made by multiplying the RS by
the inverse of the slope, 2.646. There are a total of 2089 lines identified by both our
mlethod and Bechtold et al. [3]. In addition, there are 1678 lines measured by Bechtold
et al. [3] alone. 1 Therefore, the smoothing method of §3 identified approximately
55% of those lines identified with the hand fit continua. Of the features only identified
lWe note that [3] do not measure significance as W/a(W) and they include lines with a significance
level < 3 in their lists. Only lines with SL > 3 are included in the line lists of this paper.
25
1.4
1.2
1.0
0.8
"
u)
0.6
0.6
0.4
0.2
0.0
0
1
2
3
4
5
6
7
Wobs (A)
Figure 4-1: Scatter plot of the observed RS (this paper) vs. equivalent width (as
measured by [3]). The line represents the best fit to the data.
by hand fit continua the majority are not strong features. For the lines that we did
not detect the median Wobs is 0.47
A and
the median significance level is 5.5. Just
over 85% of those lines detected solely by Bechtold et al. [3] have a significance level
below 10. For comparison, Bechtold et al. [3] measures a median Wobs of 0.93 A and
a median significance level of 11.70 for the 2089 lines identified by both methods.
Bechtold et al. [3] measures over 59% of the common lines at a significance level
greater than 10. Clearly, the features we observed are stronger than those we missed.
We conclude that our method works well when searching for strong absorption
features. While it is clear that we cannot perfectly resolve the equivalent width of an
absorption feature using our method, it is possible to create a line list which is similar
to that of the hand fit method. Also, given that virtually all the lines identified by
26
the smoothing technique are also identified by Bechtold et al. [3] (i.e. few artificial
lines are measured by the method itself), our method provides a very quick way to
search for Lyc absorbers with associated metal systems in the IGM.
4.2
Evolution of Lya Absorbers at z < 1.7
We assume that the number of absorbers in the Lya forest can be reasonably approximated by a power law in (l+z). Specifically, we characterize the distribution of lines
per unit redshift path length as:
dN
dz = Ao(1 + z)-
where Ao and y are the distribution parameters.
(4.1)
For a non-evolving population of
clouds 1/2 < y < 1 [7]. Several previous ground based studies have established that
y > 1 for z > 1.7.
However, Weymann et al. [6] and Dobrzycki et al. [7] show that
y < 1 for z < 1.7.
Assuming a "ACDM" cosmological model for the universe, it is possible to calculate the value of -y as a function of redshift. Using Equation 13.42 from Peebles
[17] we can characterize the number of absorbers per unit redshift for a non-evolving
population as:
dN
A (1 +
dz
)2
(4.2)
E(z)
where
E(z) = VIQM(1+ )3 + QA,
and Q1xM
is the mass density of the universe, and
QA
(4.3)
is the dark energy density of the
universe. If we set the right hand side of Equations 4.1 and 4.2 equal to each other.
take the log of both sides and then differentiate with respect to log(1 + z) we find:
-y(z) = 2Plugging in QM = 0.27 and
QA
3
QAI(1+ z) 3
(1
)3
Q
.
(4.4)
= 0.73 [18] we find that y = 1.41, 0.88, and 0.70 for
27
z = 0.2, 1.0, and 1.6, respectively. This suggests that over the range of redshifts we
survey y should be less than 1, in accordance with the work of [6] and [7].
Figure 4.2 shows the rest relative line strength, RSo, defined as RSxo = RSob8 /(1+
z), of all Lya lines as a function of redshift. The points marked with a cross are those
that meet the specifications of the analysis carried out in § 4.2.1. The points indicate
all the remaining SL > 3 lines in the forest. We do not include any lines blue-ward of
Ly3 emission (AQso = 1025.72) to avoid confusion with potential Ly3 absorbers. We
also remove any strong galactic absorption line features, including Si ii A1190, 1193,
1260, 1526, Si iII A1206, C
2382, 2586, 2600, Mg
ii
ii
A1334, C v A1548,1550, Al
ii
A1670, Fe
ii
A2344, 2374,
A2796, 2803, and Mg i A2853.
Unfortunately, the uniformity of the sample across all redshifts is limited by the
fact that all the observations were made by only a handful of telescope gratings. This
leads to preferential treatment of QSO sight lines where the majority of the Lya
forest is located well inside the two edges of a grating. In addition to affecting the
distribution of Lya absorption redshifts, the path length, and minimum detectable
equivalent width are also effected by the total number of sight lines associated with
each grating [6]. In the appendix, Weymann et al. [6] provide a method to correct
for the non-uniformity of the sample's coverage. They weight each component of the
empirical cumulative distribution by the minimum 4.5 a rest equivalent width at the
location of the line. This correction cannot be applied here because it depends on
the measurement of the "true" continuum, which is not what we described in § 3.
Instead, we calculate the cumulative distribution function in terms of an "effective
redshift path length" as described in [19], in which case every detected line increments
the empirical distribution function by an equal amount.
4.2.1
The Cumulative Distribution as a Function of Redshift
We establish a minimum rest relative line strength of 0.10 A for any lines that are to
be included in our analysis of the distribution of Lyo absorbers. Using the conversion
factor from above this RS threshold roughly corresponds to the 0.24 A rest equivalent
width threshold of [6] and [7]. We must also remove any lines where three times the
28
2. 0
_
I
I
I
I
I'
I
If'
,
I
i
.5
,- 1.0
C)
r
+.
Rc
.4-
+
-.
+ + ++
o (3
0.0
l4l.'
,i+
4
0.5
1.0
1.5
z (Redshift)
Figure 4-2: Rest relative line strength of all Lye lines, forest and metal systems, vs.
redshift. The crosses indicate the sample used to measure the empirical distribution
function, as described in § 4.2.1, while the dots indicate the remaining 3 r lines
detected in the forest. Only points with a RS < 2 A are shown.
signal to noise at that line is greater than the threshold minimum relative line strength
to avoid over-counting the number of lines and the total path length. This results in
the removal of many STIS lines where the S/N is high, but the three pixel window
is relatively large, meaning we would be unable to detect a 3 a line at the threshold
RS of 0.10 A. We also include all detected metal lines in our analysis. Since our
line list is in all likelihood not complete, then our identification of metal systems
would also be incomplete. To remove lines from an incomplete set of identifications
could introduce a bias into the sample by removing only the strongest metal lines.
Fortunately, Dobrzycki et al. [7]show that the inclusion of metal lines in the analysis
of Lyca number density evolution has little effect on the results.
29
As can be seen from Figure 4.2.1, the power law fit to the data proves to be remarkably good. After applying a Kolmogorov-Smirnoff test to the data in Figure 4.2.1,
we find there is a probability of 8% that a K-S statistic greater than or equal to the
observed one would arise from a random sampling of true power law distributions of
the same slope (or, PKS = 0.92). The fit is performed by a computer algorithm that
varies both y and Ao in order to maximize the K-S test statistic.
The uncertainty
in the fitting parameters is determined by assuming gaussian errors and measuring
where the K-S test statistic has fallen to e -1 / 2 of its maximum value.
1.0
0.8
0
_O
:3
' 0.6
(n
a
3 0.4
:3
0.2
0.0
0.0
0.5
1.0
1.5
z (Redshift)
Figure 4-3: The cumulative distribution of Lyo forest lines, Lya and metals lines, as
a function of redshift. The solid line shows the cumulative distribution of absorbers,
while the dashed line represents the best fit to the power law 17.3(1+z) 5 6.
30
Our best fit to the data is consistent with the non-evolving case for Lya absorbers
at low redshift. Specifically, we find that y = 0.57 + 0.16, with a corresponding value
of Ao equal to 17.3. These results are consistent with both [6] and [7], who find that y
is 0.26 i 0.22 and 0.54 i 0.21, respectively, for samples similar to our own. Dobrzycki
et al. [7] explain how the difference in sample coverage leads to a slightly lower value
of y for [6]. These results are significantly lower than ground based studies of Lya
forest lines at a redshift greater than
-
1.7. We also note that this disagrees with [20]
who found that the break in dN/dz happens near z
1.2. The largest discrepancy
between our results and those of [6] and [7] is the value of Ao. We find that A is
significantly lower in our case than Dobrzycki et al. [7], who measures Ao to be 32.7
for a sample similar to ours. We attribute this difference to largely different redshift
coverage in the two samples. We elected to remove all QSOs with a redshift less than
0.33 which means we have very little coverage of the forest at extremely low redshifts.
Weymann et al. [6] show that Ao is extremely sensitive to the value of dN/dz near
redshift 0; our limited coverage in this region is likely the cause of our relatively small
value for Ao.
4.3
Si
ii
Detection in the IGM
The search for low-ionized silicon in the IGM proves to be exceedingly difficult, because the strong lines Si
ii
A1190, 1193 and Si iii A1206 fall in the Lya forest where
the many Lya lines and the high incidence of line blending can lead to many false
positive identifications. Even the strong line at Si II A1260 quickly becomes obscured
by Lya emission before entering the forest itself as the redshift of an absorbing system
is decreased from the redshift of the QSO. Unless
Zabsorption
Zem
then Si II A1260
suffers from the same problems as the other strong silicon lines.
To avoid the problems associated with misidentification and line blending we require that our test is tuned to minimize the number of missed Si
II
systems, while at
the same time being very specific to minimize the number of false positives. We could
severely limit the number of false positives by only considering the strongest lines.
31
Then we would miss many Si Ii lines, which would prevent us from predicting their
overall frequency. False positives can be related to noise in the spectrum, or misidentified lines, especially Lya. Therefore, we use three criteria for positive identification:
line significance, redshift agreement, and the detection of at least three lines in the
system.
4.3.1
Search for Si II Lines
The following procedure is used to search for Si
ii
lines, identify possible redshift
systems, and reject unreasonable ones:
1. We search for all systems with at least two 4.5
and Si
ii
lines from the group Lya,
A1193, 1260. We also require that all three lines from this group be
detected at the 3 a level. Lines are accepted into the system if they differ from
the weighted mean
Zave
by less than 0.001. The weighted mean zave is defined
as:
Z,,e
Yi WiZi
zi Wi
=
(4.5)
where wi is the weight placed on each line:
wi
=
(4.6)
2. In systems with the three lines above we search for Lyo3, Si iII 1206 A, and C
i
1334 A in these systems. These are the most common lines associated with Si
ii
systems from previous studies [3].
3. We require that Ly:, Si
iii
1206 A, and C
ii
1334 A be 3 oadetections in order
to be included in the system.
We elect not to include systems with only Si
II
lines, because the number of false
identifications is similar to the the true number [8]. In fact, Burles & Tytler [8]
show that the number of false systems in the Lya forest dramatically decreases in the
search for three-line systems instead of two-line systems, which is why we require the
32
presence of Lya, and Si ii A1193, 1260. This is also what prevents us from searching
for any lone Si IInredshift systems. Despite the stringent criteria outlined above we
note that there is no method which will throw out all false systems and keep all the
real ones.
4.3.2
The Distribution of Si i
Table 4.1 shows the 14 systems in which we detected Si ii, as well as the system at z =
0.474 in the sight line of QSO 0454-2203. This system has a well documented redshift
system at z = 0.474 [3], and satisfies all the above criteria, with the exception of the
fact that the Si ii 1260 A line falls within the region masked for Lyc emission. The
same line is detected and identified as Si
ii
1260
A by
Bechtold et al. [3], however,
so we include this system in the table for the reader's reference. This system is not
included in any of the following analysis as it did not satisfy the necessary criteria for
identification.
Table 4.1: Si
Wavelength
ID
(A)
(A)
Zave =
Lya
1857.25 ......
Si
1961.93 ......
Si
1878.05......
Si
2077.53......
C
ii
iii
i
System Line List
Zobs
0107-0232,
1892.04 ......
ii
Zem,,
=
RSobs
a(RS)
(A)
(A)
SLa
0.728, FOS G190H
0.55644, acz= 0.00688
0.55638
0.830
0.073
11.35
1193
0.55641
0.307
0.072
4.25
1260
0.55656
0.476
0.059
8.06
1206
0.55661
0.507
0.069
7.33
0.55674
0.554
0.021
25.90
0.041
17.14
1334
0117+2118, z,,, = 1.493, FOS G270H
Zave
2634.33 ......
= 1.16711, az = 0.00336
Lya
1.16698
Continued on Next Page...
33
0.703
Table 4.1 - Continued
Wavelength
ID
(A)
( A)
A
RSobs
\
o(RS)
Zobs
tA)
(A)
SL a
Xi
2586.23......
Si II
1193
1.16731
0.352
0.057
6.19
2733.05......
Si II 1260
1.16836
0.280
0.056
4.96
2224.33......
Ly/3
1.16855
0.627
0.151
4.14
04541-2203,Zem
Zave =
= 0.533, FOS G1,90H
0.47436, oaz= 0.00095
0.47436
0.776
0.024
32.48
1193
0.47437
0.274
0.045
6.10
1857.99......
Si 111260
0.47410
1512.10 ......
Ly/3
0.47396
0.481
0.052
9.21
1778.70
......
Si iI 1206
0.47426
0.370
0.039
9.45
1967.26......
C
0.47412
0.429
0.025
17.21
1792.34 ......
Lya
1759.36......
Si
ii
1334
I
0958+5509,
Zem,,
b
b
b
= 1.750, FOS G270H
zave = 1.10679, az = 0.00120
1.10644
0.856
0.036
23.78
1193
1.10732
0.465
0.069
6.74
1260
1.10708
0.890
0.036
24.72
Si III 1206
1.10634
0.799
0.041
19.73
2560.74 ......
Lya
2514.65 ......
Si
II
2655.81 ......
Si
II
2541.30 ......
1008+1319,
zave =
Zem,
=
1.287, FOS G270H
0.90168, az = 0.00196
2311.71 ......
Lya
0.90159
1.030
0.046
22.59
2270.16......
Si II 1193
0.90244
0.376
0.081
4.64
2398.44 ......
Si
1260
0.90287
0.329
0.068
4.81
2539.30 ......
C II1
1334
0.90276
0.3121
0.064
4.85
0.042
26.83
II
1026-0045, ze,, - 1.437, FOS G270H
Zae
2791.16......
Continued
= 1.29597,az = 0.00138
Lvca
onil Next
1.29598
Page...
34
1.146
Table 4.1 - Continued
RSobs
o(RS)
Zobs
(A)
(A)
SLa
Si ii 1193
1.29590
0.325
0.074
4.42
2893..79......
Si ii 1260
1.29589
0.447
0.061
7.34
2354.36......
Ly/3
1.29531
0.850
0.074
11.51
2770.74......
Si iii 1206
1.29651
0.993
0.047
21.07
3064.28......
C
1.29614
0.434
0.063
6.87
Wavelength
ID
(A)
(A)
2739.68......
ii
1334
1115+0802, em,,= 1.718, FOS G270H
= 1.04244,
Zave
rz = 0.01052
1.04244
0.347
0.041
8.45
1193
1.04195
0.257
0.053
4.81
2574.58......
Si ii 1260
1.04263
0.297
0.039
7.64
2462.94 ......
Si III 1206
1.04139
0.591
0.039
15.31
2725.27 ......
C
1.04213
0.161
0.039
4.17
2484.92 ......
Lyc
2436.64......
Si
ii
1334
Zave =
1.39440, az = 0.00100
2911.21 ......
Lya
1.39474
0.162
0.039
4.19
2858.51 ......
Si II 1193
1.39549
0.611
0.030
20.44
3017.31......
Si ii 1260
1.39389
0.827
0.028
30.07
2887.18 ......
Si in 1206
1.39302
0.593
0.035
17.18
3196.41......
C ii 1334
1.39515
0.163
0.038
4.31
1229-0207, Zem = 1.045, FOS G190H
Zave = 0.48967, Az = 0.01501
1811.64 ......
Lya
0.49024
0.518
0.078
6.68
1778.7......
Si II 1193
0.49055
0.453
0.081
5.62
1875.,86 ......
Si iI 1260
0.48828
0.452
0.073
6.18
0.033
9.17
1248+4007, Zem= 1.030, FOS G190H
Zave
2105.87......
Lyca
= 0.73217, uz = 0.00465
0.73227
Contilued on Next Page.
35
0.305
Table 4.1 - Continued
Wavelength
ID
(A)
(A)
RSobs
7(RS)
Zobs
(A)
(A)
SL a
2067.14......
Si
ii
1193
0.73230
0.177
0.039
4.59
2183.20......
Si
ii
1260
0.73212
0.404
0.029
14.03
1775.82 ......
Ly/
0.73128
0.457
0.057
7.98
1352+0106, Zem = 1.117, FOS G190H
= 0.52395, az = 0.00117
Zave
1852.61 ......
Lyce
0.52394
0.855
0.029
29.16
1818.50 ......
Si ii 1193
0.52393
0.214
0.067
3.19
1921.02 ......
Si
1260
0.52411
0.399
0.051
7.89
1838.66 ......
Si iii 1206
0.52396
0.529
0.052
10.17
2033.86 ......
C
0.52402
0.385
0.047
8.19
ii
ii
1334
Zave = 0.66539, az = 0.00103
2024.57......
Lya
0.66539
0.814
0.026
30.94
1987.32......
Si I 1193
0.66540
0.469
0.042
11.01
2099.06 ......
Si
1260
0.66536
0.333
0.036
9.34
1708.32 .....
Ly3
0.66548
0.735
0.050
14.73
2009.54 ......
Si iii 1206
0.66559
0.541
0.037
14.62
2223.25 .....
C
0.66594
0.397
0.030
13.19
II
ii
1334
1521+1009, em = 1.321, FOS G190H
Zave = 0.67018, cz = 0.01672
2029.89 ......
Lya
1993.39 ......
Si
2104.77
......
Si
ii
0.66977
0.138
0.031
4.47
1193
0.67049
0.232
0.033
7.12
1260
0.66989
0.126
0.025
5.04
1538+4745, z,, = 0.772, FOS G190H
Zave = 0.72931,
z =
0.00020
2102.27......
Lya
0.72931
0.764
0.012
65.60
2063.90......
Si ii 1193
0.72959
0.129
0.038
3.41
Continued on Next Page...
36
Table 4.1 - Continued
RSobs
x(RS)
Zobs
(A)
(A)
Si II 1260
0.72925
0.204
0.017
12.14
1773.69......
Ly
0.72921
0.652
0.035
18.41
2086.50......
Si III 1206
0.72938
0.285
0.026
10.87
Wavelength
ID
(A)
(A)
2179.59......
--- SLa
1630+3744. Zem
, = 1.466, FOS G190H
Zave =
2050.32 ......
Lya
2013.44 ......
Si
II
2126.94 ......
Si
II
2036.74......
0.68712,a, = 0.01236
0.68658
0.283
0.041
6.98
1193
0.68730
0.289
0.049
5.86
1260
0.68748
0.265
0.036
7.39
Si III 1206
0.68814
0.410
0.044
9.25
We evaluate the evolution of Si II systems in the IGM in a method similar to the
one employed to measure the evolution of Lyacabsorbers in § 4.2. We characterize the
distribution of Si systems per unit redshift with a power law in (l+z). The sample
of lines included in the search for Si II extends beyond the Lya forest because some
systems have Si II 1260 A red-ward of Lyac emission, while others lie blue-ward of
Ly,3 emission. We restrict our search to all wavelengths greater than 1193 A in the
observed frame and less than 1260
A in
the QSO rest frame. For the cumulative
redshift path length we require that spectral coverage for 1193 A, Lya, and 1260 A
in the rest frame of the absorber exist, and that three times the signal to noise is less
than the threshold RS in each of the three corresponding pixels as well. We fit the
number of absorbers to Equation 4.1 using the method of § 4.2.1.
aSignificance level: RSb s/o(RS)
bWe measure a 3 a feature at the location of a Si II 1260 A line in a z = 0.47436 cloud, however,
this line falls in the region masked for the presence of Lya emission. We include 0454-2203 in this
table, because [3] identified a metal absorption system at z = 0.474 and because all other typical
tracers of Si II (including Si II 1190 A) are identified for this sight line. 0454-2203 is not included in
the analysis of the number of Si II absorbers per unit redshift.
37
1.0
0.8
-o
cn
0.6
o--
- 0.4
C3
0.2
0.0
0.0
0.5
1.0
1.5
z (Redshift)
Figure 4-4: The cumulative distribution of Si II lines as a function of redshift. The
solid line shows the cumulative distribution of absorbers, while the dashed line represents the best fit to the power law 2.1(1+z) - ° 5 . We note that there are large
uncertainties in the fit due to the relatively small (14) number of total systems.
The cumulative distribution, as well as the best fit to a power law distribution
are shown in Figure 4-4. We measure the fit parameters, and for the best fit case
find y = -0.5 ± 0.7 and Ao = 2.1. The K-S statistic for this fit is quite good, PKS =
0.92. The uncertainty in y is quite large, and this is believed to be the case because
the sample size is quite small (only 14 identified systems). We note that in the case
of Si
ii
redshift systems a power law of the form of Equation 4.1 may not be the best
way to parameterize the distribution of absorption systems per redshift. Our limited
sample size does not allow us to conclude one way or another about the accuracy of
the model.
38
We find the frequency of Si II in the IGM to be similar to the frequency of other
commonly occurring metal systems in the IGM. We have identified 14 systems, and a
total path length of AZtotal = 10.72. Evaluating these numbers at the mean redshift
of all the systems (z - 0.9) we find the number density of Si
< Nsiji(z = 0.9) >= 1.6 ± 0.6 .
ii
absorbers is
(4.7)
Previous measurements of the number density of other common metal systems are
also of order 1. Buries & Tytler[8] found that < N(z = 0.9) >= 1.0 i 0.6 for O vI,
while Bachall et al. [4] found < N(z = 0.3) >= 0.87 ± 0.43 for C v, and Fan [9]
found that < N(z = 0.9) >= 1.0 ± 0.25 for Mg ii. Our measurements show that the
number density of Si
ii
absorbers may be slightly higher than that of other common
metal absorbers in the low redshift regime.
39
40
Chapter 5
Conclusion
We examined the absorption properties of the Lya forest in the spectra of 204 QSOs.
We identified 832 Lya lines in these spectra, and found the evolution of hydrogen can
be described by a power law in (l+z) at low redshift. This power law is clearly flatter
than those used to describe the evolution of Lya at high redshift. This confirms the
results of previous studies of low redshift Lya evolution.
We also identified 14 Si ii systems in the spectra of these QSOs. Si
is as, if not more, common than 0 vI, C
number density of Si
ii
ii,
and Mg II at
Zae,, =
ii
absorption
0.9. We measured the
per unit redshift:
< Nsiii(z = 0.9) >= 1.6 ± 0.6 .
(5.1)
Our search for both Lya and Si ii systems was conducted with a new continuum
fitting method: we convolved the measured flux for a given sight line with a smoothing kernel designed to remove any sharp features (emission or absorption) from the
spectrum. This method only provides an approximation to the "true" continuum, but
we find that it is more than sufficient for identifying strong lines. Therefore, when
searching for strong metal absorption systems in the spectra of QSOs our method is
preferable to hand fit continua because it produces similar results in a much shorter
amount of time. We note, however, that any precise measurements of equivalent
width should be done using hand fit continua, or some other method beyond our
41
own.
Our method has some limitations, most notably near strong or sharp features in
the measured spectrum, especially Lya and C iv emission. Also, for sight lines with
lots of absorption clouds, typically those with z > 2, which were not included in the
analysis of this study, our method grossly underestimates the continuum and fails to
identify what are clearly absorption features in the spectrum. It is also clear that the
conversion from RS to equivalent width is invalid for these very strong lines. There
are some cases where our method does a good job of approximating the continuum
near emission features, however. For instance, when the rise of the feature is gradual
and it has a relatively large width, our method, especially the 5000 km s- 1 kernel,
does a good job of approximating the continuum. We note that these cases are fairly
few and far between.
We recommend that the method be used in the future for low or moderate absorption systems where the precise measurement of an absorption line's equivalent width
is of secondary importance.
Any time the measurement of the equivalent width is
important, high resolution spectra should be used, rather than the low resolution
spectra of this study. Further searches for Si ii and other common intergalactic metals should be conducted to identify what percentage of Lya absorbers are actually
hydrogen, since the current typical assumption of 100% is not correct.
42
Bibliography
[1] Lu, L., Wolfe, A. M., & Turnshek, D. A. 1991, ApJ, 367, 19
[2] Bechtold, J. 1994, ApJS, 91, 1
[3] Bechtold, J., Dobrzycki, A., Wilden, B., Morita, M., Scott, J., Dobrzycka, D., Tran, K.-V., & Aldcroft, T. A. 2002, ApJS, 140, 143
[4] Bahcall, J. N., et al. 1993, ApJS, 87, 1
[5] Bahcall, J. N., et al. 1996, ApJ, 457, 19
[6] Weymann, R. J., et al. 1998, ApJ, 506, 1
[7] Dobrzycki, A., Bechtold, J., Scott, J., & Morita, M. 2002, ApJ, 571,
654
[8] Burles, S., & Tytler, D. 1996, ApJ, 460, 584
[9] Fan, X. M. 1995, Ph.D. thesis, Columbia Univ.
[10] Keyes, C. D. et al. 1995, "FOS Instrument
Handbook",
Version 6.0,
(Baltimore: STScI)
[11] Ebbets,
D. C., Brandt,
J. C. et al. 1983, PASP, 95, 543
[12] Soderblom, D. R., et al. 1995, "GHRS Instrument Handbook", Version
6.0, (Baltimore: STScI)
[13] KinmQuijano, J., et al. 2003, "STIS Instrument Handbook", Version
7.0, (Baltimore: STScI)
43
[14] Telfer, R. C., Zheng, W., Kriss, G. A., & Davidsen, A. F. 2002, ApJ,
565, 773
[15] Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1988, ApJ, 329, L33
[16] Schlegel, D. J., Finkbeiner,
D. P., & Davis, M. 1998, ApJ, 500, 525
[17] Peebles, P. J. E. 1993, Principles
of Physical Cosmology (Princeton,
NJ: Princeton University Press), 323
[18] Spergel, D. N. et al. 2003, ApJS, 148, 175
[19] Storrie-Lombardi,
L. J., McMahon, R. G., Irwin, M. J., & Hazard, C.
1996, ApJ, 468,121
[20] Kim, T.-S., Cristiani, S., & D'Odorico,
44
S. 2001, A&A, 373, 757